3 * Common boot and setup code.
5 * Copyright (C) 2001 PPC64 Team, IBM Corp
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
15 #include <linux/config.h>
16 #include <linux/module.h>
17 #include <linux/string.h>
18 #include <linux/sched.h>
19 #include <linux/init.h>
20 #include <linux/kernel.h>
21 #include <linux/reboot.h>
22 #include <linux/delay.h>
23 #include <linux/initrd.h>
24 #include <linux/ide.h>
25 #include <linux/seq_file.h>
26 #include <linux/ioport.h>
27 #include <linux/console.h>
28 #include <linux/utsname.h>
29 #include <linux/tty.h>
30 #include <linux/root_dev.h>
31 #include <linux/notifier.h>
32 #include <linux/cpu.h>
33 #include <linux/unistd.h>
34 #include <linux/serial.h>
35 #include <linux/serial_8250.h>
38 #include <asm/processor.h>
39 #include <asm/pgtable.h>
42 #include <asm/machdep.h>
45 #include <asm/cputable.h>
46 #include <asm/sections.h>
47 #include <asm/btext.h>
48 #include <asm/nvram.h>
49 #include <asm/setup.h>
50 #include <asm/system.h>
52 #include <asm/iommu.h>
53 #include <asm/serial.h>
54 #include <asm/cache.h>
58 #include <asm/iseries/it_lp_naca.h>
59 #include <asm/firmware.h>
60 #include <asm/systemcfg.h>
65 #define DBG(fmt...) udbg_printf(fmt)
71 * Here are some early debugging facilities. You can enable one
72 * but your kernel will not boot on anything else if you do so
75 /* This one is for use on LPAR machines that support an HVC console
78 extern void udbg_init_debug_lpar(void);
79 /* This one is for use on Apple G5 machines
81 extern void udbg_init_pmac_realmode(void);
82 /* That's RTAS panel debug */
83 extern void call_rtas_display_status_delay(unsigned char c);
84 /* Here's maple real mode debug */
85 extern void udbg_init_maple_realmode(void);
87 #define EARLY_DEBUG_INIT() do {} while(0)
90 #define EARLY_DEBUG_INIT() udbg_init_debug_lpar()
91 #define EARLY_DEBUG_INIT() udbg_init_maple_realmode()
92 #define EARLY_DEBUG_INIT() udbg_init_pmac_realmode()
93 #define EARLY_DEBUG_INIT() \
94 do { udbg_putc = call_rtas_display_status_delay; } while(0)
97 /* extern void *stab; */
98 extern unsigned long klimit;
100 extern void mm_init_ppc64(void);
101 extern void stab_initialize(unsigned long stab);
102 extern void htab_initialize(void);
103 extern void early_init_devtree(void *flat_dt);
104 extern void unflatten_device_tree(void);
108 int boot_cpuid_phys = 0;
112 struct ppc64_caches ppc64_caches;
113 EXPORT_SYMBOL_GPL(ppc64_caches);
116 * These are used in binfmt_elf.c to put aux entries on the stack
117 * for each elf executable being started.
123 /* The main machine-dep calls structure
125 struct machdep_calls ppc_md;
126 EXPORT_SYMBOL(ppc_md);
128 #ifdef CONFIG_MAGIC_SYSRQ
129 unsigned long SYSRQ_KEY;
130 #endif /* CONFIG_MAGIC_SYSRQ */
133 static int ppc64_panic_event(struct notifier_block *, unsigned long, void *);
134 static struct notifier_block ppc64_panic_block = {
135 .notifier_call = ppc64_panic_event,
136 .priority = INT_MIN /* may not return; must be done last */
141 static int smt_enabled_cmdline;
143 /* Look for ibm,smt-enabled OF option */
144 static void check_smt_enabled(void)
146 struct device_node *dn;
149 /* Allow the command line to overrule the OF option */
150 if (smt_enabled_cmdline)
153 dn = of_find_node_by_path("/options");
156 smt_option = (char *)get_property(dn, "ibm,smt-enabled", NULL);
159 if (!strcmp(smt_option, "on"))
160 smt_enabled_at_boot = 1;
161 else if (!strcmp(smt_option, "off"))
162 smt_enabled_at_boot = 0;
167 /* Look for smt-enabled= cmdline option */
168 static int __init early_smt_enabled(char *p)
170 smt_enabled_cmdline = 1;
175 if (!strcmp(p, "on") || !strcmp(p, "1"))
176 smt_enabled_at_boot = 1;
177 else if (!strcmp(p, "off") || !strcmp(p, "0"))
178 smt_enabled_at_boot = 0;
182 early_param("smt-enabled", early_smt_enabled);
185 #define check_smt_enabled()
186 #endif /* CONFIG_SMP */
188 extern struct machdep_calls pSeries_md;
189 extern struct machdep_calls pmac_md;
190 extern struct machdep_calls maple_md;
191 extern struct machdep_calls cell_md;
192 extern struct machdep_calls iseries_md;
194 /* Ultimately, stuff them in an elf section like initcalls... */
195 static struct machdep_calls __initdata *machines[] = {
196 #ifdef CONFIG_PPC_PSERIES
198 #endif /* CONFIG_PPC_PSERIES */
199 #ifdef CONFIG_PPC_PMAC
201 #endif /* CONFIG_PPC_PMAC */
202 #ifdef CONFIG_PPC_MAPLE
204 #endif /* CONFIG_PPC_MAPLE */
205 #ifdef CONFIG_PPC_CELL
208 #ifdef CONFIG_PPC_ISERIES
215 * Early initialization entry point. This is called by head.S
216 * with MMU translation disabled. We rely on the "feature" of
217 * the CPU that ignores the top 2 bits of the address in real
218 * mode so we can access kernel globals normally provided we
219 * only toy with things in the RMO region. From here, we do
220 * some early parsing of the device-tree to setup out LMB
221 * data structures, and allocate & initialize the hash table
222 * and segment tables so we can start running with translation
225 * It is this function which will call the probe() callback of
226 * the various platform types and copy the matching one to the
227 * global ppc_md structure. Your platform can eventually do
228 * some very early initializations from the probe() routine, but
229 * this is not recommended, be very careful as, for example, the
230 * device-tree is not accessible via normal means at this point.
233 void __init early_setup(unsigned long dt_ptr)
235 struct paca_struct *lpaca = get_paca();
236 static struct machdep_calls **mach;
239 * Enable early debugging if any specified (see top of
244 DBG(" -> early_setup()\n");
247 * Do early initializations using the flattened device
248 * tree, like retreiving the physical memory map or
249 * calculating/retreiving the hash table size
251 early_init_devtree(__va(dt_ptr));
254 * Iterate all ppc_md structures until we find the proper
255 * one for the current machine type
257 DBG("Probing machine type for platform %x...\n",
258 systemcfg->platform);
260 for (mach = machines; *mach; mach++) {
261 if ((*mach)->probe(systemcfg->platform))
264 /* What can we do if we didn't find ? */
266 DBG("No suitable machine found !\n");
271 DBG("Found, Initializing memory management...\n");
274 * Initialize the MMU Hash table and create the linear mapping
275 * of memory. Has to be done before stab/slb initialization as
276 * this is currently where the page size encoding is obtained
281 * Initialize stab / SLB management except on iSeries
283 if (!firmware_has_feature(FW_FEATURE_ISERIES)) {
284 if (cpu_has_feature(CPU_FTR_SLB))
287 stab_initialize(lpaca->stab_real);
290 DBG(" <- early_setup()\n");
294 #if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
295 void smp_release_cpus(void)
297 extern unsigned long __secondary_hold_spinloop;
299 DBG(" -> smp_release_cpus()\n");
301 /* All secondary cpus are spinning on a common spinloop, release them
302 * all now so they can start to spin on their individual paca
303 * spinloops. For non SMP kernels, the secondary cpus never get out
304 * of the common spinloop.
305 * This is useless but harmless on iSeries, secondaries are already
306 * waiting on their paca spinloops. */
308 __secondary_hold_spinloop = 1;
311 DBG(" <- smp_release_cpus()\n");
314 #define smp_release_cpus()
315 #endif /* CONFIG_SMP || CONFIG_KEXEC */
318 * Initialize some remaining members of the ppc64_caches and systemcfg structures
319 * (at least until we get rid of them completely). This is mostly some
320 * cache informations about the CPU that will be used by cache flush
321 * routines and/or provided to userland
323 static void __init initialize_cache_info(void)
325 struct device_node *np;
326 unsigned long num_cpus = 0;
328 DBG(" -> initialize_cache_info()\n");
330 for (np = NULL; (np = of_find_node_by_type(np, "cpu"));) {
333 /* We're assuming *all* of the CPUs have the same
334 * d-cache and i-cache sizes... -Peter
337 if ( num_cpus == 1 ) {
342 /* Then read cache informations */
343 if (systemcfg->platform == PLATFORM_POWERMAC) {
344 dc = "d-cache-block-size";
345 ic = "i-cache-block-size";
347 dc = "d-cache-line-size";
348 ic = "i-cache-line-size";
352 lsize = cur_cpu_spec->dcache_bsize;
353 sizep = (u32 *)get_property(np, "d-cache-size", NULL);
356 lsizep = (u32 *) get_property(np, dc, NULL);
359 if (sizep == 0 || lsizep == 0)
360 DBG("Argh, can't find dcache properties ! "
361 "sizep: %p, lsizep: %p\n", sizep, lsizep);
363 systemcfg->dcache_size = ppc64_caches.dsize = size;
364 systemcfg->dcache_line_size =
365 ppc64_caches.dline_size = lsize;
366 ppc64_caches.log_dline_size = __ilog2(lsize);
367 ppc64_caches.dlines_per_page = PAGE_SIZE / lsize;
370 lsize = cur_cpu_spec->icache_bsize;
371 sizep = (u32 *)get_property(np, "i-cache-size", NULL);
374 lsizep = (u32 *)get_property(np, ic, NULL);
377 if (sizep == 0 || lsizep == 0)
378 DBG("Argh, can't find icache properties ! "
379 "sizep: %p, lsizep: %p\n", sizep, lsizep);
381 systemcfg->icache_size = ppc64_caches.isize = size;
382 systemcfg->icache_line_size =
383 ppc64_caches.iline_size = lsize;
384 ppc64_caches.log_iline_size = __ilog2(lsize);
385 ppc64_caches.ilines_per_page = PAGE_SIZE / lsize;
389 /* Add an eye catcher and the systemcfg layout version number */
390 strcpy(systemcfg->eye_catcher, "SYSTEMCFG:PPC64");
391 systemcfg->version.major = SYSTEMCFG_MAJOR;
392 systemcfg->version.minor = SYSTEMCFG_MINOR;
393 systemcfg->processor = mfspr(SPRN_PVR);
395 DBG(" <- initialize_cache_info()\n");
400 * Do some initial setup of the system. The parameters are those which
401 * were passed in from the bootloader.
403 void __init setup_system(void)
405 DBG(" -> setup_system()\n");
408 * Unflatten the device-tree passed by prom_init or kexec
410 unflatten_device_tree();
413 * Fill the ppc64_caches & systemcfg structures with informations
414 * retreived from the device-tree. Need to be called before
415 * finish_device_tree() since the later requires some of the
416 * informations filled up here to properly parse the interrupt
418 * It also sets up the cache line sizes which allows to call
419 * routines like flush_icache_range (used by the hash init
422 initialize_cache_info();
424 #ifdef CONFIG_PPC_RTAS
426 * Initialize RTAS if available
429 #endif /* CONFIG_PPC_RTAS */
432 * Check if we have an initrd provided via the device-tree
437 * Do some platform specific early initializations, that includes
438 * setting up the hash table pointers. It also sets up some interrupt-mapping
439 * related options that will be used by finish_device_tree()
444 * "Finish" the device-tree, that is do the actual parsing of
445 * some of the properties like the interrupt map
447 finish_device_tree();
449 #ifdef CONFIG_BOOTX_TEXT
456 #ifdef CONFIG_XMON_DEFAULT
460 * Register early console
462 register_early_udbg_console();
464 /* Save unparsed command line copy for /proc/cmdline */
465 strlcpy(saved_command_line, cmd_line, COMMAND_LINE_SIZE);
470 smp_setup_cpu_maps();
472 /* Release secondary cpus out of their spinloops at 0x60 now that
473 * we can map physical -> logical CPU ids
477 printk("Starting Linux PPC64 %s\n", system_utsname.version);
479 printk("-----------------------------------------------------\n");
480 printk("ppc64_pft_size = 0x%lx\n", ppc64_pft_size);
481 printk("ppc64_interrupt_controller = 0x%ld\n", ppc64_interrupt_controller);
482 printk("systemcfg = 0x%p\n", systemcfg);
483 printk("systemcfg->platform = 0x%x\n", systemcfg->platform);
484 printk("systemcfg->processorCount = 0x%lx\n", systemcfg->processorCount);
485 printk("systemcfg->physicalMemorySize = 0x%lx\n", systemcfg->physicalMemorySize);
486 printk("ppc64_caches.dcache_line_size = 0x%x\n",
487 ppc64_caches.dline_size);
488 printk("ppc64_caches.icache_line_size = 0x%x\n",
489 ppc64_caches.iline_size);
490 printk("htab_address = 0x%p\n", htab_address);
491 printk("htab_hash_mask = 0x%lx\n", htab_hash_mask);
492 printk("-----------------------------------------------------\n");
496 DBG(" <- setup_system()\n");
499 static int ppc64_panic_event(struct notifier_block *this,
500 unsigned long event, void *ptr)
502 ppc_md.panic((char *)ptr); /* May not return */
506 #ifdef CONFIG_IRQSTACKS
507 static void __init irqstack_early_init(void)
512 * interrupt stacks must be under 256MB, we cannot afford to take
513 * SLB misses on them.
516 softirq_ctx[i] = (struct thread_info *)
517 __va(lmb_alloc_base(THREAD_SIZE,
518 THREAD_SIZE, 0x10000000));
519 hardirq_ctx[i] = (struct thread_info *)
520 __va(lmb_alloc_base(THREAD_SIZE,
521 THREAD_SIZE, 0x10000000));
525 #define irqstack_early_init()
529 * Stack space used when we detect a bad kernel stack pointer, and
530 * early in SMP boots before relocation is enabled.
532 static void __init emergency_stack_init(void)
538 * Emergency stacks must be under 256MB, we cannot afford to take
539 * SLB misses on them. The ABI also requires them to be 128-byte
542 * Since we use these as temporary stacks during secondary CPU
543 * bringup, we need to get at them in real mode. This means they
544 * must also be within the RMO region.
546 limit = min(0x10000000UL, lmb.rmo_size);
549 paca[i].emergency_sp =
550 __va(lmb_alloc_base(HW_PAGE_SIZE, 128, limit)) + HW_PAGE_SIZE;
554 * Called from setup_arch to initialize the bitmap of available
555 * syscalls in the systemcfg page
557 void __init setup_syscall_map(void)
559 unsigned int i, count64 = 0, count32 = 0;
560 extern unsigned long *sys_call_table;
561 extern unsigned long sys_ni_syscall;
564 for (i = 0; i < __NR_syscalls; i++) {
565 if (sys_call_table[i*2] != sys_ni_syscall) {
567 systemcfg->syscall_map_64[i >> 5] |=
568 0x80000000UL >> (i & 0x1f);
570 if (sys_call_table[i*2+1] != sys_ni_syscall) {
572 systemcfg->syscall_map_32[i >> 5] |=
573 0x80000000UL >> (i & 0x1f);
576 printk(KERN_INFO "Syscall map setup, %d 32-bit and %d 64-bit syscalls\n",
581 * Called into from start_kernel, after lock_kernel has been called.
582 * Initializes bootmem, which is unsed to manage page allocation until
583 * mem_init is called.
585 void __init setup_arch(char **cmdline_p)
587 extern void do_init_bootmem(void);
589 ppc64_boot_msg(0x12, "Setup Arch");
591 *cmdline_p = cmd_line;
594 * Set cache line size based on type of cpu as a default.
595 * Systems with OF can look in the properties on the cpu node(s)
596 * for a possibly more accurate value.
598 dcache_bsize = ppc64_caches.dline_size;
599 icache_bsize = ppc64_caches.iline_size;
601 /* reboot on panic */
605 notifier_chain_register(&panic_notifier_list, &ppc64_panic_block);
607 init_mm.start_code = PAGE_OFFSET;
608 init_mm.end_code = (unsigned long) _etext;
609 init_mm.end_data = (unsigned long) _edata;
610 init_mm.brk = klimit;
612 irqstack_early_init();
613 emergency_stack_init();
617 /* set up the bootmem stuff with available memory */
621 /* initialize the syscall map in systemcfg */
624 #ifdef CONFIG_DUMMY_CONSOLE
625 conswitchp = &dummy_con;
630 /* Use the default idle loop if the platform hasn't provided one. */
631 if (NULL == ppc_md.idle_loop) {
632 ppc_md.idle_loop = default_idle;
633 printk(KERN_INFO "Using default idle loop\n");
637 ppc64_boot_msg(0x15, "Setup Done");
641 /* ToDo: do something useful if ppc_md is not yet setup. */
642 #define PPC64_LINUX_FUNCTION 0x0f000000
643 #define PPC64_IPL_MESSAGE 0xc0000000
644 #define PPC64_TERM_MESSAGE 0xb0000000
646 static void ppc64_do_msg(unsigned int src, const char *msg)
648 if (ppc_md.progress) {
651 sprintf(buf, "%08X\n", src);
652 ppc_md.progress(buf, 0);
653 snprintf(buf, 128, "%s", msg);
654 ppc_md.progress(buf, 0);
658 /* Print a boot progress message. */
659 void ppc64_boot_msg(unsigned int src, const char *msg)
661 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_IPL_MESSAGE|src, msg);
662 printk("[boot]%04x %s\n", src, msg);
665 /* Print a termination message (print only -- does not stop the kernel) */
666 void ppc64_terminate_msg(unsigned int src, const char *msg)
668 ppc64_do_msg(PPC64_LINUX_FUNCTION|PPC64_TERM_MESSAGE|src, msg);
669 printk("[terminate]%04x %s\n", src, msg);
672 #ifndef CONFIG_PPC_ISERIES
674 * This function can be used by platforms to "find" legacy serial ports.
675 * It works for "serial" nodes under an "isa" node, and will try to
676 * respect the "ibm,aix-loc" property if any. It works with up to 8
680 #define MAX_LEGACY_SERIAL_PORTS 8
681 static struct plat_serial8250_port serial_ports[MAX_LEGACY_SERIAL_PORTS+1];
682 static unsigned int old_serial_count;
684 void __init generic_find_legacy_serial_ports(u64 *physport,
685 unsigned int *default_speed)
687 struct device_node *np;
690 struct isa_reg_property {
695 struct pci_reg_property {
696 struct pci_address addr;
701 DBG(" -> generic_find_legacy_serial_port()\n");
707 np = of_find_node_by_path("/");
711 /* First fill our array */
712 for (np = NULL; (np = of_find_node_by_type(np, "serial"));) {
713 struct device_node *isa, *pci;
714 struct isa_reg_property *reg;
715 unsigned long phys_size, addr_size, io_base;
717 u32 *interrupts, *clk, *spd;
719 int index, rlen, rentsize;
721 /* Ok, first check if it's under an "isa" parent */
722 isa = of_get_parent(np);
723 if (!isa || strcmp(isa->name, "isa")) {
724 DBG("%s: no isa parent found\n", np->full_name);
728 /* Now look for an "ibm,aix-loc" property that gives us ordering
731 typep = (char *)get_property(np, "ibm,aix-loc", NULL);
733 /* Get the ISA port number */
734 reg = (struct isa_reg_property *)get_property(np, "reg", NULL);
737 /* We assume the interrupt number isn't translated ... */
738 interrupts = (u32 *)get_property(np, "interrupts", NULL);
739 /* get clock freq. if present */
740 clk = (u32 *)get_property(np, "clock-frequency", NULL);
741 /* get default speed if present */
742 spd = (u32 *)get_property(np, "current-speed", NULL);
743 /* Default to locate at end of array */
744 index = old_serial_count; /* end of the array by default */
746 /* If we have a location index, then use it */
747 if (typep && *typep == 'S') {
748 index = simple_strtol(typep+1, NULL, 0) - 1;
749 /* if index is out of range, use end of array instead */
750 if (index >= MAX_LEGACY_SERIAL_PORTS)
751 index = old_serial_count;
752 /* if our index is still out of range, that mean that
753 * array is full, we could scan for a free slot but that
754 * make little sense to bother, just skip the port
756 if (index >= MAX_LEGACY_SERIAL_PORTS)
758 if (index >= old_serial_count)
759 old_serial_count = index + 1;
760 /* Check if there is a port who already claimed our slot */
761 if (serial_ports[index].iobase != 0) {
762 /* if we still have some room, move it, else override */
763 if (old_serial_count < MAX_LEGACY_SERIAL_PORTS) {
764 DBG("Moved legacy port %d -> %d\n", index,
766 serial_ports[old_serial_count++] =
769 DBG("Replacing legacy port %d\n", index);
773 if (index >= MAX_LEGACY_SERIAL_PORTS)
775 if (index >= old_serial_count)
776 old_serial_count = index + 1;
778 /* Now fill the entry */
779 memset(&serial_ports[index], 0, sizeof(struct plat_serial8250_port));
780 serial_ports[index].uartclk = clk ? *clk : BASE_BAUD * 16;
781 serial_ports[index].iobase = reg->address;
782 serial_ports[index].irq = interrupts ? interrupts[0] : 0;
783 serial_ports[index].flags = ASYNC_BOOT_AUTOCONF;
785 DBG("Added legacy port, index: %d, port: %x, irq: %d, clk: %d\n",
787 serial_ports[index].iobase,
788 serial_ports[index].irq,
789 serial_ports[index].uartclk);
791 /* Get phys address of IO reg for port 1 */
795 pci = of_get_parent(isa);
797 DBG("%s: no pci parent found\n", np->full_name);
801 rangesp = (u32 *)get_property(pci, "ranges", &rlen);
802 if (rangesp == NULL) {
808 /* we need the #size-cells of the PCI bridge node itself */
810 sizeprop = (u32 *)get_property(pci, "#size-cells", NULL);
811 if (sizeprop != NULL)
812 phys_size = *sizeprop;
813 /* we need the parent #addr-cells */
814 addr_size = prom_n_addr_cells(pci);
815 rentsize = 3 + addr_size + phys_size;
817 for (;rlen >= rentsize; rlen -= rentsize,rangesp += rentsize) {
818 if (((rangesp[0] >> 24) & 0x3) != 1)
819 continue; /* not IO space */
820 io_base = rangesp[3];
822 io_base = (io_base << 32) | rangesp[4];
825 *physport = io_base + reg->address;
826 if (default_speed && spd)
827 *default_speed = *spd;
834 DBG(" <- generic_find_legacy_serial_port()\n");
837 static struct platform_device serial_device = {
838 .name = "serial8250",
839 .id = PLAT8250_DEV_PLATFORM,
841 .platform_data = serial_ports,
845 static int __init serial_dev_init(void)
847 return platform_device_register(&serial_device);
849 arch_initcall(serial_dev_init);
851 #endif /* CONFIG_PPC_ISERIES */
853 int check_legacy_ioport(unsigned long base_port)
855 if (ppc_md.check_legacy_ioport == NULL)
857 return ppc_md.check_legacy_ioport(base_port);
859 EXPORT_SYMBOL(check_legacy_ioport);